User:Gaz Lloyd/Disassembly

The RuneScape Wiki wants your disassembly data!

Hey there. If you're training Invention, you're probably disassembling a lot of things in bulk. All we want you to do is make a note of what you get and send us your data.

You can submit data here. If you need more information, see below.

Gathering data

The data gathering process is very simple. Here's a step-by-step guide:

1. Choose and obtain the item you're going to disassemble.
   • The current minimum sample size we're accepting is 50, but the larger the better. If you can do 1000s of an item, that's fantastic.
   • For things that disassemble more than one at a time (e.g. runes), submit your total amount disassembled, not the amount of disassemble actions (i.e. 3000 runes not 60 x (50 runes)). You can use the feedback section at the end of the form to specify that you did (x lots of y item for xy total) if you need to.
   • It doesn't really matter what item you're doing, we want it all.
2. Check and note down your current junk chance reduction.
   • You can find this by following these steps: open the invention skill guide, change the filter to everything, scroll down until you find the last instance of "junk chance reduction" that is marked "researched" - image help
   • If you don't know it, it doesn't really matter (see below)
3. Analyse your item and note down what materials it gives
4. Open your invention pouch and note down the current amount of all of the materials (from the analysis) that you currently have. Include special materials and junk.
   • If you miss out junk, it doesn't really matter (see below)
5. Do the disassembly.
   • Don't disassemble anything else in this process; only this item until you do the next step
   • Do not use B.A.N.K. Standers during the disassembly - they'll throw off material counts and (possibly) corrupt your entire data set to unusability
   • If you are using the invention master cape, make sure to keep it on for the entire disassembly process
6. Open your invention pouch again and note down the numbers again.
7. Take the original set of numbers away from the second set to find how much stuff you obtained from the disassembly.
8. Submit this data using the form!

If you're doing large quantities of items (e.g. a few thousand), it is advisable to break this down into sets (of e.g. 1,000 items each) and note down totals at the end of each set. This can help find and mitigate transcription/calculation errors later should your numbers come out wrong - we'd only have to scrap 1,000 of your sample, instead of all 15,000.

Required data summary

As a quick summary, this is what we need:

• (Exact) name of item
• Number of that item (at least 50 please!)
• Your current junk chance reduction
• If you are wearing the invention master cape
• Names/amounts of each of the materials you obtained by disassembling the items, including special materials and junk

Thanks in advance!

Junk

The inclusion of junk in this form is mostly a sanity check. It doesn't actually factor in to the real calculations; we can function well enough without. As such, if you don't know your junk amount or reduction, submit anyway. You can set junk to 0 or whatever.

If you do have an issue with junk, having not recorded it or whatever, please note that in the 'any other feedback' question at the end of the form.

Bulk bulk data

If you've done a lot of different items (and can't be bothered to use the form), you can put it into a spreadsheet and send me a link to that. I'll input the data directly to save time.

You can send the link to me via:

• My talk page here
• My twitter
• Messaging me on reddit
• If I'm online:
  • Discord
  • Or in-game if you want to chat

Make sure you format your spreadsheet in a way that makes it clear which number is which component. If you're doing lots in a category, make sheets for each category. If you're doing one in each category, maybe you'd like to format it like the results of the form (that makes it easy for me!).

Things to get data for

Invention master cape

The invention master cape perk is generally understood, but it would be good to fully confirm it - but we need some absurd amount of data to do that!

Normal categories

Needs data

These should be easy to get data for, since they're all tradeable and relatively cheap

• Clockwork

Needs data, but hard to get

At time of writing, the disassembly categories that we most don't have data for are:

• Whips
• Defenders
• Defender pieces

Extra is only mildly useful

Low priority but might have some use for extra data.

• Faceted glass

Extra data isn't needed

Low priority categories we have enough data to get a meaningful distribution out of, but more data is always welcome to help confirm distributions:

• Two-handed crossbows
• Two-handed swords
• Arrows
• Ashed incense sticks
• Ashes
• Battleaxes
• Bolts
• Bones
• Claws
• Clothing
• Crossbows and stocks
• Daggers
• Darts
• Divination products
• Food and herbs
• Gems
• Glass items
• Halberds
• Hatchets
• Herbed incense sticks
• Hybrid armour
• Jewellery
• Throwing knives
• Logs
• Longswords
• Maces
• Mauls
• Melee armour
• Magic off-hand weapons
• Pickaxes
• Summoning pouches
• Ranged armour
• Scimitars
• Summoning scrolls
• Seeds
• Shieldbows
• Shields
• Shortbows
• Spears
• Staves
• Swords
• Throwing axes
• Wands
• Warhammers

No need

These only give one material anyway, so we definitely don't need them:

• Ore
• Charms

And these have been provided by shauny, so we know the true distribution for them:

• Bladed salvage
• Blunt salvage
• Plated salvage
• Spiky salvage
• Gizmos and relics
• Potions
• Planks, bars, leather
• Crystal tool
• Cannonballs
• Runes
• Magic armour

Use and results

This section assumes you've read and understood Disassemble#Mechanics

The form submissions are visible here. If you notice something off, you can comment on the relevant cell.

Anatomy of the sheet

The spreadsheet is essentially in 4 parts.

• The initial sheet, named Form Responses 1 (with a cyan label) is where all the responses are added, both automatically by the form and manually by me
  Additionally, I add the 'category' column to each row to define which category the item is from, and if needed, the 'don't use' column if the data seems off (or is outdated)
• The next sheet, named Summary (with a green label), which finds out how many non-junk materials have been submitted for each category, excluding those to not include
  Those with a total greater than the borderline value (15,000) are highlighted; those which are already put on to the wiki have the 'Analysed' column set to true
• The next set are two variants of essentially the same thing: sheets for each category, which filter the responses into only those of one category, and perform analysis on them
  • The ones which have enough data, and have been put on to the wiki, are marked with a purple label and generally pushed toward the end of the sheet list
  • Incomplete ones have no colour on their label and are between the completed ones (purple) and the summary sheet (green)

Basic error-checking

In the form responses sheet, starting in column Y, some basic error checking is carried out. The total of the supplied material counts and junk are compared to what is expected from the sample. If the difference is more than a few and there is no explanatory comment, the data will be set to "don't use" and is excluded from all analysis. If the error is allowed, the reason is left in the justification column. Further checking is done by eye in each of the specific sheets.

Analysis

The analysis performed on the data is on the far right of each category sheet (column Y onwards). Through spreadsheet magic it calculates the total and percentage of each material, as well as a confidence interval for each (95% [see cell M8 in Summary sheet], calculated in the same way as our charm logs - the Wilson score interval for binomial proportion confidence intervals).

From this, once a large enough sample of materials has been found (I generally say 15,000 non-junk, though sometimes more are needed), I take the most likely true values from two main assumptions: for the majority of categories, all percentages are integers (and, of course, add up to 100%); and the materials are listed in the Analyse window in decreasing probability order.

Implementation

Once enough data has been obtained, and the percentages found, they are added to Module:Disassembly material calculator/data. This then automatically filters through and allows the display of the relevant items in the relevant material calculators.

Once enough categories have data, the percentages will also be shown in the main disassembly template on every disassemblable item page.